1.2
- ensure_precision now makes exception catching obsolete for automatic precision adaptation.
Please be careful when using this routine and read the doc of precsion_control::is_autoadapting.
- benchmark_n now publicly provides routines for assessing the run-time behavior of functions.
Similarly configurationtools_n is a set of routines for optimal parameter determination.
- new, faster pi, log(2) and atan formulae (log(2) & Pi of Bailey et al.)
- improved handling of limit cases in trigo routines
- stricter enforcing of inversion-of-exp property of log
- fixed documentation of real
- minor enhancements of basic arithmetic
1.1
- The new naming scheme has been further applied, in particular to number types.
- real_c::mantissa() now allows for read-write access
(faster low-level operations without any hacks)
- real_c::accumulate provides a fast method for implementation of series expansions.
- A critical bug in sqrt has been identified and removed.
1.0
- A new naming scheme is introduced with complex classes being suffixed with "_c" and simpler
types with "_t". Getter-routines are loosing their "get_" prefix.
- Tidied-up source tree
- MPI interface (optional)
- faster incrementation (++-operator)
- New version numbering policy
0.5
- improved source tree layout
- sign representation simplified
- ++ and -- operators added to "real" interface.
- New simple functions that come in very handy, e.g. for comparison of absolute values.
- significant improvement of log performance.
- improvements to trigonometric functions
- bugfixes: boolean arithmetic, multiplication
0.4.2:
- Bugfix in Karatsuba multiplication
0.4.1:
- Added "Changelog"
- Added benchmarking module
- configure now requires APMATH_TESTPROGRAM instead of __APMATH_CONFIGURATION_PROGRAM
- Added "min_precision" & "max_precision" to precision_control
- Smarter memory management in multiplication (speed up).
- "random_number_generator" is renamed to "random".
- "div_2_pow/mul_2_pow" are more robust.